Thanks for your patience while we dug into this.

We looked at the host your backend was running on around the time you reported, and there was a noisy-neighbor issue. The host became partially I/O-saturated and drove up I/O wait for your workload and a handful of others sharing it. This was isolated to that single host and it was not a quota or migration change on our side, and the host has since recovered to a healthy baseline after a short period.

We are sorry for the disruption this caused. We are working on systemic changes to prevent this kind of host contention from impacting workloads going forward.

Are you still seeing degraded performance today?

From what I can tell, things should be back to baseline.

No we still see degraded performance and we are still above 4 times the baseline.

Taking a look, I can't see anything on our side that would have caused this outside of a short period on may 18th (around 3 hours) when one of your API replicas landed on a contended host. Even then, it was only one of those replicas that would have been affected. Can you provide any more details on exactly what you're seeing that points at CPU I/O Wait / contention?

As I said, outside of that short window, you should be back to a healthy basline re: CPU performance